The system currently check validity of the import file and mapped fields in the initial steps, takes you to the screen to review possible duplicates (Step 4) (and allows you to select merge/replace/ignore etc for each entry or in bulk) and then has a further validation screen (step 5) that picks up any errors (for example blank data rows, missing last name field).  Once an error is found and this step, you have to go back to the start, upload your import file again and go through the possible duplicates screen process (it does not remember previous settings).

The system should perform a  full validation of the data file BEFORE it makes you go through the laborious process of reviewing/updating possible duplicates.

Alternatively, if errors are found that prevent the file being imported, it should allow you to edit or exclude those lines so as not to hold up the whole process.